00:24:27 · Srinivas Narayanan, Aditya Agarwal
[00:24:27] S0: Yeah. I would say, to be blunt, I think OpenAI was behind in coding for a while. Right? I think Anthropic had an early lead in coding. They invested in it, and I think they did a better job at it than OpenAI did in the past. But I think I would say in the last several months, it's not clear that there is such a big lead. I think coding... Codex is fantastic product. The team has done a fantastic job. So I I would say most of the reason why there was a... OpenAI was behind was because the pace at which coding products took off and found product market fit was very unprecedented. And even a small lead became a very big lead for Anthropic, but I don't know that the latest numbers. But I actually think technology wise and product wise, I would say Codex is fantastic, and I use it all the time. So, hopefully, you know, OpenAI has caught up recently as well. On the consumer side, yeah, I mean, ChatGPT is a fantastic product. We continue to use that. I think the last thing is enterprise adoption we talked about. That is just going slower, like I said, for all the reasons, but I think that will take more time to diffuse into the world. [00:25:30] S1: Sure. It does... From the outside and when you look at the competition, it does feel like a fistfight. Right? Of course, there are the American model companies. There are the Chinese model companies and a bunch of action which is happening across the world. And in that kind of a context, when you're looking at internally strategic choices, how do you even think about what to go after?
00:25:30 · Aditya Agarwal, Srinivas Narayanan
[00:25:30] S1: Sure. It does... From the outside and when you look at the competition, it does feel like a fistfight. Right? Of course, there are the American model companies. There are the Chinese model companies and a bunch of action which is happening across the world. And in that kind of a context, when you're looking at internally strategic choices, how do you even think about what to go after? [00:25:48] S0: I think for OpenAI, at least when I was there, it's a very clear mission. Right? We are here to build AGI, and it's a very clear mission for how do we advance the research to build models that are capable at solving tasks that are economically useful and increasingly also in scientific research. Right? But the progress to AGI was very, very clear. And, yes, you know, it's gonna be incredibly competitive because these models and these products are incredibly useful. But I don't think we kinda distracted ourselves too much. Yes. There are gonna be other models. There's gonna be other competitors, but I don't think you can, like, let that distract you too much. I mean, you pay attention to what they're doing if there's something you can learn. And, you know, of course, you learn that and try to take good insights from whatever you can learn, but I don't think you get too distracted by what other people are doing. I think we had a very good mission. We're continuing... The the company is still continuing that mission. And I think the market is also extremely large. So Yeah. I think there'll be lots of good companies and winners in the space. Sure. You know, one
00:25:48 · Srinivas Narayanan, Aditya Agarwal
[00:25:48] S0: I think for OpenAI, at least when I was there, it's a very clear mission. Right? We are here to build AGI, and it's a very clear mission for how do we advance the research to build models that are capable at solving tasks that are economically useful and increasingly also in scientific research. Right? But the progress to AGI was very, very clear. And, yes, you know, it's gonna be incredibly competitive because these models and these products are incredibly useful. But I don't think we kinda distracted ourselves too much. Yes. There are gonna be other models. There's gonna be other competitors, but I don't think you can, like, let that distract you too much. I mean, you pay attention to what they're doing if there's something you can learn. And, you know, of course, you learn that and try to take good insights from whatever you can learn, but I don't think you get too distracted by what other people are doing. I think we had a very good mission. We're continuing... The the company is still continuing that mission. And I think the market is also extremely large. So Yeah. I think there'll be lots of good companies and winners in the space. Sure. You know, one [00:26:56] S1: of the conversations which have been there is that intelligence will get commoditized. Right? And even the best large... Best labs will possibly have a limited value capture. Right? I have a personal counter to that, but I'm just curious on your personal belief there. I think it's true that people in
00:27:14 · Srinivas Narayanan
[00:27:14] S0: the world have more choice today than they ever had in the past. Right? And the open models are actually quite fantastic at doing lots of tasks. And I think that kind of choice is actually good for companies, for startups, everyone. Yeah. I also believe that there is always a demand for the frontier models. And today, there are only a very small number of companies who are able to produce these any... At any point in time, the frontier models, and there is incredible demand for them because I think there's... We always find new use cases for being at the frontier. You know, six months ago, it was coding. Maybe some of the coding... You can get good coding models from open models as well. But then if you want to do leading math or scientific research, it's not clear that you can do it from right. So there's always new applications that emerge for being at the frontier, and I I continue to believe that will also happen. The other thing to say is that just being open doesn't mean it's cheaper. Right? I think there is enormous advantages to scale for a lot of the frontier labs. And especially when you operate at scale, you can have efficiencies. And so you might still produce models with similar intelligence but which may be even more cost effective. Having said that, I think there's also a big market for frontier... Sorry, for open models. You know, it gives you control. It gives you customization, and I think that's good for a lot of cases. So I don't think that it'll only be a or b. I think it'll be both. And the last thing I would say is that, you know, we are in a world where we are incredibly compute constrained. And so, you know, compute and having access to compute is also a huge, huge strategic advantage. And today, the frontier labs seem to have a lot of that. And so I think that will also continue to play to that advantage.
00:28:57 · Aditya Agarwal
[00:28:57] S1: 100%. Yeah. I know I was joking with someone that at times, teams in these labs, when they're complaining about the lack of compute, that seems to be more than what, as a country, that we have in India, right? So interesting times. Let's talk about the future a bit here, right? Of course, language is a scaffolding for human thought. But intelligence needs various other inputs. Have a bunch of talk around world models and models for specific purposes, models for science, and a bunch of other things. How do you view that part of, you know, emergent thoughts around some of the new ways of how possibly people should be interacting with, yeah, with both processing and availability of intelligence on those access.
00:29:46 · Srinivas Narayanan
[00:29:46] S0: No. I mean, the the opportunity to take intelligence and actually make useful things in the real world is is really different now compared to six months ago. Yeah. It's very exciting that, you know, the forefront of what you can imagine has moved from just purely digital software to how does intelligence interact with the real world. Right? So you're seeing enormous amount of innovation happening in robotics, in AI interacting with the physical world, in being able to do science, whether it's material science, whether it's life science. You know, math is already happening to some extent. Right? So I think you're gonna see that quite a bit. The cost of writing software has, you know, gone down dramatically as a result of all these models. So it is sort of like push the frontier of what people are thinking about from purely being in a digital world to actually being, okay. Now how can we use this in the physical world? Physical world starts... Has constraints. You know? How fast can we get verifiable data back and verifiable rewards back because, you know, that's not happening at the speed of light anymore? So that creates different challenges, but that is the opportunity. You know? You need new challenges, and I think there's a lot of innovation happening in how do you accelerate the feedback loop and how do we do that. So I think it's quite exciting.